Everyone is eager to improve; and this is why many of us are turning towards AI recommendations for modifying sentence structures, grammar, and vocabulary. Can it be that instead of improving our personal phrasing and fluency in writing and thinking, by using AI text editor software, we neutralize our characteristic way of wording ideas? If everyone is using this software, can it be that, eventually, all digital text will sound and look the same?
I started using software such as ‘grammar.ly’, yet only recently I started wondering about the implications of using AI software to improve text documents. How much of the text still belongs to me after using ‘grammar.ly’ to re-phrase my paragraph a number of times? Can I even still say I wrote it? Is it my text – or the AI’s?
A person’s writing is, in theory, unique, and this is why certain traits of a person can be distilled from their wording and style (1). This is called stylometry and with it, it is possible to link different documents to the same person, their demographic, even age groups and sex (2). The usage of deep learning allowed stylometry models to become even more sophisticated, so much so that these models now allow digital forensics to analyse harassment emails in order to determine whether those have been written by the same author (3)(4)(5).
However, what if that author uses a generative AI model, even ChatGPT, to write the message? To what conclusion would digital forensics come to, could they still bypass the AI shield and connect the text to the human author? What does it mean when we hide behind the AI?
I am certain that, despite the higher use of AI in digital writing, we could be able to advance other technologies in order to bypass the ‘AI shield’. However, the higher use of AI in digital writing might mean the loss of many writing styles, including human writing styles.
The issue we are now facing is the lack of motivation to write long texts and articles. Even now, many people make use of AI to automate writing their emails, assignments and even private messages.
What happens to academia when AI –think of the AI-Bing search– has full access to the internet and is able to source and cite properly? Imagine being able to prompt your research question and receive a master thesis within minutes – written by an AI. The way I see it, this is the direction we are headed towards; from simply using ‘grammar.ly’, or advancing with ChatGPT and others.
Sure, AI is a great tool to automate, especially administrative tasks, but when it comes to expressing our ideas, we should exercise our human potential, despite the ease of AI usage. Otherwise, we will run the risk of losing our edge.
1: https://ieeexplore.ieee.org/abstract/document/6234430
2: https://www.researchgate.net/publication/221298006_Stylometric_Analysis_of_Bloggers%27_Age_and_Gender
3: Tweedie, F. J., Singh, S., & Holmes, D. I. (1996). Neural Network Applications in Stylometry: The “Federalist Papers.” Computers and the Humanities, 30(1), 1–10. http://www.jstor.org/stable/30204514)
Thank you for sharing this information with us. I have the experience with grammarly, because sometimes I myself think that after adapting my text according to the recommendations, I feel like the text does not belong to me anymore. However, by using this tool for a longer period of time I started seeing some improvements in terms of my grammar. I really like that you touched upon the stylometry because after ChatGPT was available a lot of students I know started using it to generate ideas and even write long texts. Consequently, we might end up losing creativity and willingness to write extensive papers (good point about thesis). I think nowadays it is really possible to write the whole thesis by using AI content generators. But the problem is that we might end up losing our personality, creativity and critical thinking when it comes to academic papers/reports which might yield a significant damage to the educational system. To conclude I totally agree with your statement “Sure, AI is a great tool to automate, especially administrative tasks, but when it comes to expressing our ideas, we should exercise our human potential, despite the ease of AI usage. Otherwise, we will run the risk of losing our edge.”
This is an intriguing topic, indeed I agree with you that we lose a lot of our creative ability when we choose to engage with Generative AI for day-to-day requirements. When I’m typing this comment on the blog, I’m able to independently process my thoughts without being affected by word and phrase suggestions, whereas if I was typing this on other document editors such as Word or Google doc, or even Gmail, I keep getting suggestions on how I should complete my sentence. Before my mind is able to think for itself, I’m being suggested the most commonly used words and once you take note of it, it is difficult to come up with another word on your own, or even think differently.
Initially when this technology was introduced, I had gone the extra mile to turn it off, however now it is present everywhere by default and you have to take the effort to actually turn it off from each of these applications. There is no explicit permission asked from the user before this technology started becoming omnipresent. Grammar.ly maybe more advanced that these inbuilt tools but Grammar.ly is something you install consciously whereas there are these Generative AI tools in your life that you never explicitly gave consent to. I believe these word/phrase suggestions could lead to lower linguistic and stylistic diversity in writing. Your article is spot on to recognize how this is even more amplified with the use of ChatGPT. Thank you for sharing your thoughts on this topic and prompting a discussion on this issue.
This article was quick to catch my attention due to: (1) my regular use of grammar.ly, and (2) my interest in the use of generative AI in academics. As I read this article, I reflected on the subconscious use of this tool and how it has impacted my individual writing style. What I realised is that while my grammer use has improved slightly (As Vlad also mentioned), I find that I also reject a lot of suggestions for alternative words or phrasing.
I also quite agree with your thoughts on how people may be facing a lack of motivation in the face of generative AI, however, I believe that the way in which its being used varies a lot depends on the educational level but also certain skills such as critical thinking. I have observed that extent of the generated text that one accepts (as is) comes partially from the individual’s ability to critically evaluate it, look at the evidence presented and derive sound conclusions. This is in the scope of full text generation more so than brainstorming.
As you noted, it could be that with more advancements, the evaluation may be less needed in terms of the correctness of the information. However, I wholeheartedly believe that when it comes to idea expression or in more nuanced domains, we should not rely on generative AI to avoid running the risk of converging/unoriginal ideas and knowledge, and ‘digital text sounding the same’.
Very interesting post, Bijan!